Wi-Fi Speed Tests: So F*ucking What!

I was reading my twitter feed on the way home tonight and it was abuzz about a new Miercom report that Cisco had paid for that showed their new Aironet 3702i AP vs. an Aruba Networks AP-225. Miercom is well known by most WLAN pros that I know in that they generally give you paid results, pay enough and you product will look wonderful vs. the competition! Once I got off the train and into my car, one of my favorite Metallica tracks began blasting. ‘So f*cking what! … Well who cares, who cares what you do…’ I began thinking about this report and AP speed tests in general and that I really didn’t care about them. The song really was just what I felt about this and the previous Miercom report that Aruba themselves had commissioned.

Lets be honest here. There’s always gonna be someone claiming their AP is the fastest thing since butter. There are very few good wireless engineers who I know that take speed alone as the single biggest factor in selecting an AP for their customers. In my experience what matters the most is the budget of the customer. Every time I discuss with people in detail what they are really looking for it comes down to what the main use case is. I carefully explain to them that the report they’ve been reading about vendor X’s super fast APs really doesn’t relate to their environment. The reason for this is pretty much everyone’s environment is different and it also changes from day to day as things change when RF interacts with all the different objects and people in that environment. The top speed of vendor X’s AP is also going to vary, sometimes wildly.

I’ve used an Aruba AP-225 for several months in my office and it works really fantastically well for me. I’m pretty sure that my friends who have Cisco APs that they use daily in their offices also have the same experience (or any other vendor’s AP for that matter). What 802.11ac brings for folks and why it’s important is that it increases capacity overall. That won’t matter a dime to someone who has a low number of clients and who’s users are generally not doing a lot on the wireless network. So what’s my advice to you who is looking at potentially upgrading and trying to choose between all these vendor’s products? Start with what your needs are today, do an honest extrapolation of how many more clients and what version of 802.11 they will potentially be using then use that as your basic metric. Oh, and hire a good WLAN engineer to help you sort it all out if it just seems confusing to you. They will save you money in the long run because you will buy what you need and they will design it so it works properly for your environment.

Gigabit Wi-Fi Design – More Devices, More Capacity, More Problems.

Fluke 802.11ac

When I first started learning how to deploy Wi-Fi networks I really only had a vague knowledge of how it all really worked. In those days, I relied heavily on friends who were experts to give me a basic idea of how Access Points should be deployed. It was in a way much simpler to design WLANs as well, as usually you were only concerned with covering all the places people usually would use their laptops in a building. This requirement could be ‘we just want conference rooms’ or ‘we just want coverage in the executive suites as everyone else has desktops’. My first major deployment was a college, however, and their requirements were to have coverage everywhere students went with their laptops. This required us to design coverage for over 70 buildings and additional outdoor areas which was a huge undertaking for me. I learned a lot about proper RF design from doing this.

The major change we have seen in the last few years is that the main way people on any device access the network is wireless. We have mobile phones, tablets, laptops and all sorts of other devices using the WLAN and this means that RF design has changed from just making sure coverage is good to making sure the network performs well everywhere. This is a huge change in emphasis from ‘nice to have’ to ‘mission critical’. It’s simply not good enough anymore to throw up APs and hope the WLAN performs acceptably.

This was emphasized for me in the first session at Wireless Field Day 5 with the Fluke Networks AirMagnet Product division presentation. They wanted to talk about the updates to the Survey Pro product they were working on to support 802.11ac, the upcoming Wi-Fi standard. For an excellent overview of the impact of 802.11ac I highly recommend Andrew vonNagy’s Revolution Wi-Fi blog series on 802.11ac Gigabit Wi-Fi. For this blog, I am going to concentrate on the updates to AirMagnet Survey Pro.

First of all I will state that a large part of the driving force behind Gigabit Wi-Fi being marketed to consumers (almost too much hype) is the explosion in wireless devices that are being used. This has resulted in the consumerization of IT and major headaches for most IT departments in handing the load on their WLANs. It is still, however very early days with these devices as the standard is not yet ratified and there is a large amount of inconsistent performance in client devices being sold. Even just that the first phase client devices support 2×2:2 (that’s two transmit and receive antennas and 2 spatial streams) means those devices can potentially do 867Mbps over the air and this could saturate USB2 connections. So from the client side, USB3 is needed and I would say that the biggest gain will be when device manufacturers integrate 802.11ac into their products with PCI-Express cards.

The uneven client performance will eventually be fixed by driver or chip updates but this is just one of the factors to consider. Using older 802.11a clients will substantially slow the network down and there is additional considerations if clients support 1, 2 or 3 spatial streams. From the AP side of things in the WLAN contributing factors are channel allocations, denser deployments to support high modulation rates (256-QAM) and many others. Basically as the environment has gotten more complex and variable it means less predictable performance and that the PHY rate is no longer an accurate predictor of performance. That’s without even adding in the challenges of trying to use 80MHz channels.

AirMagnet Survey Pro is attempting to meet these challenges head on by bringing into their tool features needed to more clearly measure client performance. The reality here is that even for experienced users of Survey Pro, your methodology will incorporate some kind of measuring upstream and downstream performance testing to optimize client networks. Survey Pro incorporates iPerf so that performance data can be gathered as part of an active survey. Fluke believes that doing onsite validation surveys with 802.11ac and incorporating performance testing into that is the way to verify that the network will perform to client specifications.

Other additions to Survey Pro that are more specific to 802.11ac are new heatmap views that show primary and secondary channel overlap. This is particularly needed for 80MHz channels as channel planning will need to account for interference of secondary channels (CCI). Another view shows where APs are getting the new 256-QAM rate, which should give a good overview of the performance dropoff as clients move away from the AP. For those of you with a Gold Support contract for Survey Pro you can go to this page to sign up for the beta and see for yourself some of these features in action. You will of course need a supported 802.11ac adapter.

It will be challenging for WLAN designs as 802.11ac begins to rollout. More thought will need to be put into performance testing by validation engineers to make sure that the design meet’s customer criteria. I expect that client drivers will eventually even out as 802.11ac becomes a more mainstream technology. Unfortunately we will probably have to go through another round of this with MU-MIMO in Phase 2, but more on that in a later blog.

Cloudy with a chance of meat donuts

I just got back from an amazing week at Wireless Field Day 5. Being a big deal on the Internet came up and I made sure to push my case that my mum thinks I’m a big deal on the Internet too. I expect though that all mums (or moms if you prefer) think their kids are a big deal. In any case, I happen to have the pleasure of doing these events with a bunch of wireless engineers who believe I’m a big enough deal that I keep getting invited back. I think it’s probably something to do with me being opinionated enough to ask tough questions and be honest when I’m wrong or don’t have all the information.

We had a great bunch of presenters this time around and I applaud every one of you for showing up in front of a bunch of wireless geeks when you didn’t really know what to expect from us. I will, however, make a general statement that ignoring questions doesn’t make them go away in front of such a crowd, it just makes those asking more determined to find out what it is you’re hiding. Two of the presenters, Xirrus and Meru Networks have controversial products, which naturally elicited from us a bunch of questions about how they deal with the issues we see with their approach. Although I cannot say that either one really satisfied the requests for information, the approach taken by the presenters could not have been more different. Suffice to say, this was my reaction to the plain ignoring the question and pleading ignorance that Meru Networks choose to take (as recorded by Blake Krone);

RAWR

First of all I would like to say that Xirrus founder and executive chairman Dirk Gates showed an amazing amount of technical knowledge for a CEO and went out of his way to try his best to answer the questions being thrown at him by the delegates. He did default a few time to not diving quite as deeply as we would like but he at least promised to bring back engineers to answer our questions next time. It was also apparent with the demonstration portion of the talk that the arrays do work well for large venues that need to deal with 10’s of thousands of clients. I think that more than anything, the willingness shown by Dirk and his team to be up front and open about their tech and to try to answer the details we were asking for was admirable. I still have some reservations surrounding other client use cases, but I accept that for their main market, large venues, it works well.

Meru Networks, however, chose a different tack. That was to gloss over the issues with Co-Channel Interference (CCI) in their product and plead ignorance about the technical details when the person being asked was clearly a technical manager. This just annoys technical people and makes us even more determined to either a) find out what you are hiding or b) conclude that really your product has a basic issue that prevents it’s widespread adoption in the enterprise. As the presentation went on, it became clear that Meru was going to ignore questions about CCI and so I can only conclude that the latter is the correct conclusion.

This matches what I have thought about Meru for quite some time now, there is a basic, physical limitation to their Single Channel Architecture (SCA) that means it cannot scale well. This will likely be exacerbated by 802.11ac when it comes out. I believe that Meru will have more and more difficulty in convincing customers of the benefits of their products and I will continue to see Meru as I have seen them up until now, which is them being replaced because their solution just doesn’t perform well in a high capacity environment that modern enterprises need.

If I can give advice to any company who is planning on attending a wireless field day, it would be first, put on your fire pants because you will get asked tough questions from the delegates and second, it would be bring along your top engineers to answer questions. The companies that have most impressed us were the one’s who were not shy to bring forward their technical genius who had a deep understanding of their products. We as a group are deeply passionate about wireless networks and RF (and in my case, wireless security) and because of this we have a depth of knowledge you won’t get from any other audience.